Module Dependency

In [1]:
%matplotlib inline

import numpy as np
import skimage as ski
from skimage import exposure, io
from skimage.morphology import disk, square
from skimage.filters import rank, median, gaussian
from skimage.color import rgb2hsv, hsv2rgb, rgb2gray, gray2rgb
import matplotlib.pyplot as plt

plt.rcParams['figure.figsize'] = (20.0, 16.0)

Read Image

In [2]:
image = io.imread('./data/input2.bmp')
plt.imshow(image)
Out[2]:
<matplotlib.image.AxesImage at 0x7f76a9e71a90>

Median Filter with kernel size [10, 10]

In [3]:
for i in range(3):
    image[:, :, i] = median(image[:, :, i], square(10))
plt.imshow(image)
Out[3]:
<matplotlib.image.AxesImage at 0x7f76a9df1ac8>

Plot RGB Distribution

In [4]:
f, axarr = plt.subplots(3)
axarr[0].hist(image[:, :, 0].ravel(), bins=256, color='r')
axarr[1].hist(image[:, :, 1].ravel(), bins=256, color='g')
axarr[2].hist(image[:, :, 2].ravel(), bins=256, color='b')
plt.show()

Replace V Channel with Blue Channel/255

In [5]:
image_hsv = rgb2hsv(image)
# image_hsv[:, :, 2] = exposure.equalize_hist(image[:, :, 2])
image_hsv[:, :, 2] = image[:, :, 2]/255
image = hsv2rgb(image_hsv)
plt.imshow(image)
Out[5]:
<matplotlib.image.AxesImage at 0x7f76a11c6748>

Perform Global Histogram Equalization and Show V Channel Histogram

In [6]:
image_hsv = rgb2hsv(image)
image_hsv[:, :, 2] = exposure.equalize_hist(image_hsv[:, :, 2])
plt.hist(image_hsv[:, :, 2].ravel(), bins=256)
plt.show()

Add some Saturation

In [7]:
image_hsv[:, :, 1] *= 2
image_hsv[:, :, 1][image_hsv[:, :, 1] > 1] = 1
image = hsv2rgb(image_hsv)
plt.imshow(image)
Out[7]:
<matplotlib.image.AxesImage at 0x7f76a7c30160>

Rescale Intensity to Increase Constrast

In [8]:
# image[:, :, 0] = exposure.rescale_intensity(image[:, :, 0], in_range=(55, 255))
for i in range(3):
    image[:, :, i] = exposure.rescale_intensity(image[:, :, i], in_range=(0.1, 1))
plt.imshow(image)
Out[8]:
<matplotlib.image.AxesImage at 0x7f76a7c0d198>

Write Image

In [9]:
io.imsave('./data/output2.bmp', image)

/home/dennis/anaconda3/lib/python3.6/site-packages/skimage/util/dtype.py:122: UserWarning: Possible precision loss when converting from float64 to uint8
  .format(dtypeobj_in, dtypeobj_out))